Contact us

Author: olewave

News
Exploring Wav2vec 2.0 fine-tuning for improved speech emotion recognition
Tutorials

์Œ์„ฑ ๊ฐ์ • ์ธ์‹ ๊ฐœ์„ ์„ ์œ„ํ•œ Wav2vec 2.0 ๋ฏธ์„ธ ์กฐ์ • ํƒ์ƒ‰

์Œ์„ฑ ๊ฐ์ • ์ธ์‹์€ ์˜ค๋žซ๋™์•ˆ ์ž‘์€ ๋ ˆ์ด๋ธ”๋œ ๋ฐ์ดํ„ฐ์…‹๊ณผ ์†์œผ๋กœ ๋งŒ๋“  ์Œํ–ฅ ํŠน์„ฑ์— ๊ฐ‡ํ˜€ ์žˆ์—ˆ์Šต๋‹ˆ๋‹ค.

12์›” 4, 2021ยท 1 min read
Joint Unsupervised and Supervised Training for Multilingual ASR
Tutorials

๋‹ค๊ตญ์–ด ASR์„ ์œ„ํ•œ ๊ณต๋™ ๋น„์ง€๋„ํ•™์Šต ๋ฐ ์ง€๋„ํ•™์Šต ํ›ˆ๋ จ

์‚ฌ์ „ ํ›ˆ๋ จ ํ›„ ๋ฏธ์„ธ ์กฐ์ •์ด ๋‹ค๊ตญ์–ด ASR์˜ ๊ธฐ๋ณธ ๋ ˆ์‹œํ”ผ๊ฐ€ ๋˜์—ˆ์ง€๋งŒ, ๋งŽ์€ ์ž ์žฌ๋ ฅ์„ ๋‚จ๊ธฐ๊ณ  ์žˆ์Šต๋‹ˆ๋‹ค: ๋‘ ๋‹จ๊ณ„๊ฐ€ ์†์‹ค์„ ๊ณต์œ ํ•˜์ง€ ์•Š์œผ๋ฉฐ, ์ง€๋„ ๋‹จ๊ณ„๊ฐ€ ์œ ์šฉํ•œ ์ž์ฒด ์ง€๋„ ๊ตฌ์กฐ๋ฅผ ์กฐ์šฉํžˆ ์ œ๊ฑฐํ•  ์ˆ˜ ์žˆ์Šต๋‹ˆ๋‹ค

12์›” 4, 2021ยท 1 min read
Improving Speech Recognition Accuracy of Local POI Using Geographical Models
Tutorials

์ง€๋ฆฌ์  ๋ชจ๋ธ์„ ์‚ฌ์šฉํ•˜์—ฌ ์ง€์—ญ POI ์Œ์„ฑ ์ธ์‹ ์ •ํ™•๋„ ๊ฐœ์„ ํ•˜๊ธฐ

์Œ์„ฑ ์–ด์‹œ์Šคํ„ดํŠธ์— ์ž‘์€ ์ง€์—ญ ์‹๋‹น์œผ๋กœ ์ด๋™ํ•˜๋„๋ก ์š”์ฒญํ•˜๋ฉด ASR์ด ์‹คํŒจํ•˜๋Š” ์ง€์ ์„ ๋น ๋ฅด๊ฒŒ ๋ฐœ๊ฒฌํ•  ์ˆ˜ ์žˆ์Šต๋‹ˆ๋‹ค โ€” LM ํ•™์Šต ์ฝ”ํผ์Šค์— ๋‚˜ํƒ€๋‚˜์ง€ ์•Š์•˜๋˜ ๋กฑํ…Œ์ผ POI ์ด๋ฆ„๋“ค์ž…๋‹ˆ๋‹ค.

11์›” 19, 2021ยท 1 min read
UniSpeech-SAT : Universal Speech Representation Learning with Speaker Aware Pre-Training
Tutorials

UniSpeech-SAT: ํ™”์ž ์ธ์‹ ์‚ฌ์ „ ํ•™์Šต์„ ํ†ตํ•œ ๋ฒ”์šฉ ์Œ์„ฑ ํ‘œํ˜„ ํ•™์Šต

์ž๊ธฐ ์ง€๋„ ์Œ์„ฑ ๋ชจ๋ธ์€ ์Œ์„ฑ ๋‚ด์šฉ์„ ์•„๋ฆ„๋‹ต๊ฒŒ ํ•™์Šตํ•˜๋Š” ๊ฒฝํ–ฅ์ด ์žˆ์ง€๋งŒ ํ™”์ž ์‹ ์›์€ ๋‚˜์ค‘์— ์ƒ๊ฐํ•˜๋Š” ๊ฒฝํ–ฅ์ด ์žˆ์Šต๋‹ˆ๋‹ค. ์ด๋Š” ASR์—๋Š” ์ข‹์ง€๋งŒ ํ™”์ž ๊ฒ€์ฆ, ํ™”์ž ๋ถ„ํ•  ๋˜๋Š” ์Œ์„ฑ ๋ณต์ œ์—๋Š” ํ˜•ํŽธ์—†์Šต๋‹ˆ๋‹ค.

11์›” 13, 2021ยท 1 min read
RefineGAN: Universally Generating Waveform Better than Ground Truth with Highly Accurate Pitch and
Tutorials

RefineGAN: ๋†’์€ ์ •ํ™•๋„์˜ ํ”ผ์น˜์™€ ํ•จ๊ป˜ ๊ธฐ๋ณธ ์ง„์‹ค๋ณด๋‹ค ๋” ๋‚˜์€ ํŒŒํ˜•์„ ๋ณดํŽธ์ ์œผ๋กœ ์ƒ์„ฑ

"๊ธฐ๋ณธ ์ง„์‹ค๋ณด๋‹ค ๋‚˜์Œ"์€ ๋ˆˆ์„ ๊ตด๋ฆฌ๊ฑฐ๋‚˜ ์žฌ์ƒํ•˜๊ฒŒ ๋งŒ๋“œ๋Š” ์ข…๋ฅ˜์˜ ์ฃผ์žฅ์ž…๋‹ˆ๋‹ค.

11์›” 5, 2021ยท 1 min read
SNRi Target Training for Joint Speech Enhancement and Recognition
Tutorials

๊ณต๋™ ์Œ์„ฑ ๊ฐœ์„  ๋ฐ ์ธ์‹์„ ์œ„ํ•œ SNRi ๋ชฉํ‘œ ํ›ˆ๋ จ

์Œ์„ฑ ๊ฐœ์„  ๋ฐ ASR์˜ ๊ณต๋™ ํ›ˆ๋ จ์€ ๋ถ„๋ช…ํ•ด ๋ณด์ž…๋‹ˆ๋‹ค. ๋…ธ์ด์ฆˆ ์ œ๊ฑฐ๊ธฐ๋ฅผ ํ†ตํ•ด ์ธ์‹ ์†์‹ค์„ ์—ญ์ „ํŒŒํ•˜๋ฉด ๋ฉ๋‹ˆ๋‹ค. ํ•˜์ง€๋งŒ ์‹ค์ œ๋กœ๋Š” ๊ณต๊ฒฉ์ ์ธ ๋…ธ์ด์ฆˆ ์ œ๊ฑฐ๊ฐ€ ์Œํ–ฅ ๋ชจ๋ธ์ด ์˜์กดํ•˜๋Š” ๊ธฐ๋Šฅ์„ ์ž์ฃผ ์™œ๊ณกํ•˜๊ธฐ ๋•Œ๋ฌธ์— ๋ณต์žกํ•ฉ๋‹ˆ๋‹ค.

11์›” 5, 2021ยท 1 min read
BigSSL: Exploring the Frontier of Large-Scale Semi-Supervised Learning for Automatic Speech Recog
Tutorials

BigSSL: ์ž๋™ ์Œ์„ฑ ์ธ์‹์„ ์œ„ํ•œ ๋Œ€๊ทœ๋ชจ ์ค€์ง€๋„ ํ•™์Šต์˜ ๊ฒฝ๊ณ„ ํƒ์ƒ‰

100M ๋งค๊ฐœ๋ณ€์ˆ˜์—์„œ ์ตœ์ฒจ๋‹จ์„ ์ด๊ธด ์ค€์ง€๋„ ASR ๋ ˆ์‹œํ”ผ๋ฅผ ๊ฐ€์ ธ๋‹ค๊ฐ€ ๊ณ„์† ํ™•์žฅํ•˜๋ฉด ์–ด๋–ป๊ฒŒ ๋ ๊นŒ์š”?

10์›” 29, 2021ยท 1 min read